Abstract
Neural networks have an important presence in geography and the environment, where they are used to classify and extract various characteristics in images. Detecting features in sea ice imagery is extremely relevant to today's society as sea ice continues to melt due to rising temperatures. In this project, we have researched different types of parallelism in order to gain a better understanding of how to make the training process more efficient. We have also worked with both AWS and Google Collab, using both CPUs and GPUs, to train the classification algorithm to differentiate types of clouds (rainy/non rainy) which are the same color. We also trained the algorithm to classify types of sea ice (thick, thin) based on how transparent the ice is. Having an accurate algorithm that can extract characteristics in sea ice is important to measure the rate at which sea ice is melting in an easy manner and to correctly back up claims about the impact of climate change.
How to Cite:
, C., , Y., , D., , K. & , J.,
(2022) “Spatiotemporal Innovation Testbed”,
Journal of Student-Scientists' Research 3.
doi: https://doi.org/10.13021/jssr2021.3254
Files
Downloads are not available for this article.